Plasma-assisted physical-vapor deposition (PVD) using various sputtering target materials is a commonly used thin-film fabrication technology for manufacturing of semiconductor, data storage, flat-panel display, and photovoltaic devices. Plasma sputtering is the most important PVD fabrication technique. In semiconductor integrated circuit (IC) manufacturing, PVD processes are utilized to deposit contact/via barrier (e.g., TiN or TaN) and glue (e.g., Ti) layers as well as metallization materials (e.g., Al or Cu). State-of-the-art semiconductor technologies developed for producing high-performance logic devices such as microprocessor chips employ four to six levels of metal interconnects. Current salicided CMOS technologies with four levels of interconnects may use (a) one PVD step (e.g., Ti deposition) to form self-aligned silicide (i.e., salicide), (b) two PVD steps to deposit Ti and TiN layers at the contact level, (c) one PVD step such as a PVD reflow process to form the contact plug and the first level of interconnect (typically Al, plus 1% Si, and 0.5% Cu), (d) one PVD step to deposit an Anti-Reflection Coating (ARC) layer to facilitate the interconnect patterning process, and (e) between six to nine additional PVD process steps to form the three additional interconnect levels and associated via liner and barrier layers as well as the ARC layers. This process description illustrates that advanced semiconductor IC manufacturing technologies with multiple levels of interconnect often require numerous PVD process steps, some with stringent step coverage and bottom coverage requirements in order to produce continuous coverage over high aspect-ratio contact and via structures.
The performance, reliability, and yield of the multilevel interconnects strongly influence the reliability and manufacturing yield of semiconductor chips such as high-performance microprocessors. As a result, PVD fabrication processes play a significant role in semiconductor integrated circuits, since they influence all the significant multilevel interconnect performance and reliability parameters.
Existing commercial PVD technologies usually employ DC magnetron sputtering in vacuum chambers. A typical commercial PVD equipment includes a single-substrate (single-wafer) vacuum process chamber (preferably designed as a cluster tool module), a temperature-controlled chuck (with the option to apply electrical bias power to the chuck) to hold the substrate, and a sputter target (or magnetron cathode) that contains the desired material. DC magnetron plasma excitation (with DC power levels up to 10 kW to 20 kW) is usually used for sputter deposition of electrically conductive materials such as Al, Ti, Co, and TiN. RF magnetron sputtering is usually used for sputter deposition of electrically insulating (or resistive) materials. RF diode sputtering (as opposed to magnetron PVD sputtering) is the preferred choice for sputter deposition of some magnetic materials and insulating materials for applications such as thin-film head fabrication.
Each of these PVD methods generates a plasma from an inert plasma gas such as Ar and sustain the plasma near the target area. The target material atoms or molecules are then sputtered from the target surface and deposited on the device substrate. Sputter etching of the target occurs due to the impact of energetic Argon ion species. During the sputtering process, the sputtered species (mostly neutrals) are emitted within the vacuum chamber plasma environment over a wide range of spatial angles and a portion of the sputtered flux deposits on the device substrate. Other sputtering processes, such as reactive sputtering processes, use nitrogen or oxygen or another reactive gas, instead of, or in addition to, an inert gas within the vacuum chamber. Reactive magnetron sputtering processes that deposit TiN layers from elemental Ti targets illustrate an example of this technique.
In general, the flux of the sputter atoms or species that the PVD target material emits has a relatively broad angular distribution. Similarly, the sputter flux of species arriving at the substrate surface has a relatively broad distribution angle. This broad distribution angle does not present a problem in applications involving substrates without high aspect ratio features or with relatively minor surface topography features.
However, some semiconductor device manufacturing applications involve substrates with high aspect ratio features. These applications require some degree of spatial filtering or collimation for the sputter species. A broad angular distribution of the PVD flux implies poor collimation or a low degree of collimation, whereas a narrow angular distribution relative to the perpendicular axis indicates a higher degree of PVD collimation. For instance, semiconductor interconnect applications require collimated sputtering for deposition of the contact and via liner/glue and barrier layers (e.g. Ti/TiN) when using high aspect ratio (for instance, on the order of .gtoreq.3:1) contacts and vias due to the bottom coverage and sidewall coverage requirements. For a contact/via hole of width (or diameter) W and height H, the following parameters can be defined:
A.R. .DELTA.H/W (aspect ratio) PA1 Bottom Coverage .DELTA.t.sub.b /d PA1 Sidewall Coverage .DELTA.t.sub.s /d (stop coverage)
where d is the thickness of sputtered material layer on extended flat top surfaces, t.sub.b is the sputtered material thickness at the bottom of the hole, and t.sub.s is the thickness of the sputtered material on the hole sidewall at mid height.
In conventional PVD processes without any built-in sputtering collimation feature, the bottom coverage and sidewall coverage of the sputter deposited material degrades significantly as the microstructure aspect ratio increases. This degradation becomes increasingly and rapidly worse for microstructure aspect ratios of greater than 3:1. As a result, for applications requiring good bottom coverage (e.g., .gtoreq.25%) and sidewall coverage (.gtoreq.50%), existing PVD technologies use a collimator plate placed between the PVD target or cathode assembly and the substrate inside the vacuum chamber as shown in FIG. 1 (prior art). This collimator plate provides a physical collimation technique.
The collimator plate, usually made of aluminum or titanium, consists of an array of circular or hexagonal (honeycomb) closely packed holes (see FIG. 1(b) that typically have an aspect ratio of 1.5:1 or higher. The collimator plate operates as a spatial filter to reduce the angular distribution of the sputter flux species arriving at the substrate. The degree of collimation increases (thus decreasing the angular distribution of the sputter flux species at the substrate) as the collimator plate hole aspect ratio increases. Although physical collimation via a fixed collimator plate is extensively practiced in commercial PVD equipment for semiconductor contact/via glue/liner and barrier layer depositions, it has a number of drawbacks and disadvantages.
First of all, a trade-off exists between the sputter deposition rate (throughput) and the degree of collimation. In other words, higher degrees of collimation require collimators with higher aspect ratios, resulting in a reduced deposition rate. This trade-off in physical collimated PVD systems presents a significant throughput limitation. Secondly, the collimator becomes gradually coated with the sputtering material. This can result in particulate generation within the PVD vacuum chamber. Moreover, the coating creates a non-uniform blockage of the collimator holes that can result in long-term drift of the process uniformity. These problems lead to the requirement for frequent removal and cleaning of the collimator plate. The need for frequent cleaning and maintenance of the collimator has a negative impact on the overall equipment uptime and product throughput. Finally, a fixed collimator plate does not allow for variable (adjustable in real time) and controlled degrees of collimation, since the degree of collimation is fixed by the aspect ratio of the collimator. Some processes require multi-step depositions with and without PVD collimation in order to maximize the process throughput while, at the same time, establishing optimal hole or trench filling capability. These processes must use multiple PVD stations (including PVD stations or modules with collimation and those without collimation) on a cluster tool which results in an increased cost-of-ownership for the multi-step process.
Another prior art PVD collimation method is the "natural collimation" or "long-through" collimation method. The natural collimation method relies on providing a relatively large spacing (as compared to the sputtering cathode and substrate diameters) between the target and the substrate. In this collimation technique, higher collimation ratios require larger target to substrate spacing values. Like the physical collimation method with a collimator plate, the natural collimation technique also suffers from a trade-off between the degree of collimation and the deposition rate. Moreover, a higher collimation ratio in a natural collimation method results in a higher PVD chamber volume, requiring a higher pumping speed pump package.
Thus, there is a need for an ultraclean PVD collimation technique which does not generate particulates and does not require regular and frequent replacement or cleaning of the PVD chamber components.
Moreover, there is a need for an improved PVD collimation that does not suffer from a trade-off between the deposition rate and the degree of PVD collimation.
There is also a need for an improved collimation technique which provides a capability for adjustable and controlled collimation (programmable electrical collimation) of the PVD process in real time. The collimation control capability will be established via a controlled and adjustable electrical parameter (substrate bias, ionization source RF power, or both of these parameters) with or without a capability to adjust the target-to-substrate spacing.
There is also a need for a PVD technology that can perform high throughput processing of substrate using high rate deposition with enhanced processing capability without requiring higher PVD chamber volume.
There is also a need for a PVD technology that can deposit material layers with excellent thickness control and process control capability for a wide range of material layer thicknesses from very thin (down to 10 .ANG.) to relatively thick layers.
There is also a need for a PVD technology that provides a capability to sustain a stable and reproducible plasma medium for a wide range of PVD cathode power levels (e.g., from zero to the highest power level on the cathode).